Implement Structured Output with Pydantic Model

[KUNDAN1334]

Solved #8

[Copilot]

Pull Request Overview

This PR implements structured output support for the ChatGradient model using Pydantic models, enabling the model to return validated, type-safe responses conforming to predefined schemas.

Key Changes:

• Added with_structured_output() method to ChatGradient that accepts Pydantic models, TypedDict, or JSON schemas
• Implemented JSON extraction and parsing utilities with error handling
• Added comprehensive unit tests covering basic Pydantic output, validation errors, and raw mode

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 6 comments.

File	Description
langchain_gradient/chat_models.py	Implemented with_structured_output() method and helper functions for parsing JSON from model responses
tests/unit_tests/test_structured_output.py	Added unit tests for structured output functionality with Pydantic models
docs/structured_output_examples.md	Added documentation and usage examples for structured output feature

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The regex pattern r\"``````\" will only match exactly 6 backticks in a row. Based on the test mock responses which use triple backticks (), this pattern should be `r\"(.+?)```"` to correctly extract JSON from markdown code blocks.

Suggested change

	pattern = r"``````"
	pattern = r"```(.+?)```"

[Copilot]

The prompt template contains 6 backticks but the regex pattern expects 6 backticks exactly. This creates an inconsistency. The template should use triple backticks (```) to match standard markdown code block syntax, or the regex pattern in _extract_and_parse_json needs to be updated accordingly.

[Copilot]

The instruction says 'wrap your JSON output in `````` tags' but standard markdown code blocks use triple backticks (```). This mismatch will cause parsing to fail. Change to triple backticks or update the regex pattern to match 6 backticks.

Suggested change

	"Always wrap your JSON output in `````` tags."),
	"Always wrap your JSON output in ``` tags."),

[Copilot]

The file ends with 'undefined' which appears to be an unintended placeholder or error. This line should be removed.

Suggested change

undefined

[Copilot]

The code block is missing proper markdown formatting. It should be wrapped in triple backticks with the language identifier (```python) to render correctly as a code block.

Suggested change

	from pydantic import BaseModel, Field
	from langchain_gradient import ChatGradient
	class Person(BaseModel):
	name: str = Field(description="The person's name")
	age: int = Field(description="The person's age")
	email: str = Field(description="Email address")
	llm = ChatGradient(model="llama3.3-70b-instruct")
	structured_llm = llm.with_structured_output(Person)
	response = structured_llm.invoke(
	"Create a person named John, age 30, email john@example.com"
	)
	print(response)
	Output: Person(name='John', age=30, email='john@example.com')
	## Multiple Objects
	from typing import List
	class PersonList(BaseModel):
	people: List[Person]
	structured_llm = llm.with_structured_output(PersonList)
	response = structured_llm.invoke("Create 3 people with different names")
	print(response.people) # List of Person objects
	## Error Handling
	Get raw output and errors
	structured_llm = llm.with_structured_output(Person, include_raw=True)
	result = structured_llm.invoke("Create a person")
	if result["parsing_error"]:
	print(f"Error: {result['parsing_error']}")
	print(f"Raw: {result['raw'].content}")
	else:
	print(f"Parsed: {result['parsed']}")
	```python
	from pydantic import BaseModel, Field
	from langchain_gradient import ChatGradient
	class Person(BaseModel):
	name: str = Field(description="The person's name")
	age: int = Field(description="The person's age")
	email: str = Field(description="Email address")
	llm = ChatGradient(model="llama3.3-70b-instruct")
	structured_llm = llm.with_structured_output(Person)
	response = structured_llm.invoke(
	"Create a person named John, age 30, email john@example.com"
	)
	print(response)

Output: Person(name='John', age=30, email='john@example.com')

Multiple Objects

from typing import List

class PersonList(BaseModel):
    people: List[Person]

structured_llm = llm.with_structured_output(PersonList)
response = structured_llm.invoke("Create 3 people with different names")
print(response.people) # List of Person objects

Error Handling

Get raw output and errors

structured_llm = llm.with_structured_output(Person, include_raw=True)

result = structured_llm.invoke("Create a person")

if result["parsing_error"]:
    print(f"Error: {result['parsing_error']}")
    print(f"Raw: {result['raw'].content}")
else:
    print(f"Parsed: {result['parsed']}")

[Copilot]

This line should be formatted as a comment or descriptive text (e.g., '# Get raw output and errors') rather than appearing as standalone text outside a code block.

Suggested change

	Get raw output and errors
	**Get raw output and errors**